diff --git a/apps/api/src/integrations/pl24/pl24.service.ts b/apps/api/src/integrations/pl24/pl24.service.ts
index ca1f545..bb97b93 100644
--- a/apps/api/src/integrations/pl24/pl24.service.ts
+++ b/apps/api/src/integrations/pl24/pl24.service.ts
@@ -66,13 +66,20 @@ const LEGACY_ARCH_SOURCE_TAG: Record<string, string> = {
  * single spelling matches both "Model yılı" and "MODEL YILI".
  */
 export function normalizeLabel(label: string): string {
-  return label
-    .toLocaleLowerCase("tr")
-    .normalize("NFD")
-    .replace(/[\u0300-\u036f]/g, "")
-    .replace(/ı/g, "i")
-    .replace(/[\s/]+/g, "_")
-    .trim();
+  return (
+    label
+      .toLocaleLowerCase("tr")
+      .normalize("NFD")
+      // \p{M} (all combining marks) rather than the U+0300–U+036F range: the range
+      // is a character class that can also match a base character followed by a
+      // combining one, which biome flags as misleading. NFD has already split every
+      // accent into its own mark, so matching marks directly is both correct and
+      // broader (Turkish, Latin-Extended, anything PL24 sends).
+      .replace(/\p{M}/gu, "")
+      .replace(/ı/g, "i")
+      .replace(/[\s/]+/g, "_")
+      .trim()
+  );
 }
 
 /**